1 /* 2 * linux/include/linux/jbd2.h 3 * 4 * Written by Stephen C. Tweedie <sct@redhat.com> 5 * 6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved 7 * 8 * This file is part of the Linux kernel and is made available under 9 * the terms of the GNU General Public License, version 2, or at your 10 * option, any later version, incorporated herein by reference. 11 * 12 * Definitions for transaction data structures for the buffer cache 13 * filesystem journaling support. 14 */ 15 16 #ifndef _LINUX_JBD2_H 17 #define _LINUX_JBD2_H 18 19 /* Allow this file to be included directly into e2fsprogs */ 20 #ifndef __KERNEL__ 21 #include "jfs_compat.h" 22 #define JBD2_DEBUG 23 #else 24 25 #include <linux/types.h> 26 #include <linux/buffer_head.h> 27 #include <linux/journal-head.h> 28 #include <linux/stddef.h> 29 #include <linux/mutex.h> 30 #include <linux/timer.h> 31 #include <linux/slab.h> 32 #include <linux/bit_spinlock.h> 33 #include <crypto/hash.h> 34 #endif 35 36 #define journal_oom_retry 1 37 38 /* 39 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds 40 * certain classes of error which can occur due to failed IOs. Under 41 * normal use we want ext4 to continue after such errors, because 42 * hardware _can_ fail, but for debugging purposes when running tests on 43 * known-good hardware we may want to trap these errors. 44 */ 45 #undef JBD2_PARANOID_IOFAIL 46 47 /* 48 * The default maximum commit age, in seconds. 49 */ 50 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5 51 52 #ifdef CONFIG_JBD2_DEBUG 53 /* 54 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal 55 * consistency checks. By default we don't do this unless 56 * CONFIG_JBD2_DEBUG is on. 57 */ 58 #define JBD2_EXPENSIVE_CHECKING 59 extern ushort jbd2_journal_enable_debug; 60 void __jbd2_debug(int level, const char *file, const char *func, 61 unsigned int line, const char *fmt, ...); 62 63 #define jbd_debug(n, fmt, a...) \ 64 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a) 65 #else 66 #define jbd_debug(n, fmt, a...) /**/ 67 #endif 68 69 extern void *jbd2_alloc(size_t size, gfp_t flags); 70 extern void jbd2_free(void *ptr, size_t size); 71 72 #define JBD2_MIN_JOURNAL_BLOCKS 1024 73 74 #ifdef __KERNEL__ 75 76 /** 77 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. 78 * 79 * All filesystem modifications made by the process go 80 * through this handle. Recursive operations (such as quota operations) 81 * are gathered into a single update. 82 * 83 * The buffer credits field is used to account for journaled buffers 84 * being modified by the running process. To ensure that there is 85 * enough log space for all outstanding operations, we need to limit the 86 * number of outstanding buffers possible at any time. When the 87 * operation completes, any buffer credits not used are credited back to 88 * the transaction, so that at all times we know how many buffers the 89 * outstanding updates on a transaction might possibly touch. 90 * 91 * This is an opaque datatype. 92 **/ 93 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ 94 95 96 /** 97 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. 98 * 99 * journal_t is linked to from the fs superblock structure. 100 * 101 * We use the journal_t to keep track of all outstanding transaction 102 * activity on the filesystem, and to manage the state of the log 103 * writing process. 104 * 105 * This is an opaque datatype. 106 **/ 107 typedef struct journal_s journal_t; /* Journal control structure */ 108 #endif 109 110 /* 111 * Internal structures used by the logging mechanism: 112 */ 113 114 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ 115 116 /* 117 * On-disk structures 118 */ 119 120 /* 121 * Descriptor block types: 122 */ 123 124 #define JBD2_DESCRIPTOR_BLOCK 1 125 #define JBD2_COMMIT_BLOCK 2 126 #define JBD2_SUPERBLOCK_V1 3 127 #define JBD2_SUPERBLOCK_V2 4 128 #define JBD2_REVOKE_BLOCK 5 129 130 /* 131 * Standard header for all descriptor blocks: 132 */ 133 typedef struct journal_header_s 134 { 135 __be32 h_magic; 136 __be32 h_blocktype; 137 __be32 h_sequence; 138 } journal_header_t; 139 140 /* 141 * Checksum types. 142 */ 143 #define JBD2_CRC32_CHKSUM 1 144 #define JBD2_MD5_CHKSUM 2 145 #define JBD2_SHA1_CHKSUM 3 146 #define JBD2_CRC32C_CHKSUM 4 147 148 #define JBD2_CRC32_CHKSUM_SIZE 4 149 150 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) 151 /* 152 * Commit block header for storing transactional checksums: 153 * 154 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* 155 * fields are used to store a checksum of the descriptor and data blocks. 156 * 157 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum 158 * field is used to store crc32c(uuid+commit_block). Each journal metadata 159 * block gets its own checksum, and data block checksums are stored in 160 * journal_block_tag (in the descriptor). The other h_chksum* fields are 161 * not used. 162 * 163 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses 164 * journal_block_tag3_t to store a full 32-bit checksum. Everything else 165 * is the same as v2. 166 * 167 * Checksum v1, v2, and v3 are mutually exclusive features. 168 */ 169 struct commit_header { 170 __be32 h_magic; 171 __be32 h_blocktype; 172 __be32 h_sequence; 173 unsigned char h_chksum_type; 174 unsigned char h_chksum_size; 175 unsigned char h_padding[2]; 176 __be32 h_chksum[JBD2_CHECKSUM_BYTES]; 177 __be64 h_commit_sec; 178 __be32 h_commit_nsec; 179 }; 180 181 /* 182 * The block tag: used to describe a single buffer in the journal. 183 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this 184 * raw struct shouldn't be used for pointer math or sizeof() - use 185 * journal_tag_bytes(journal) instead to compute this. 186 */ 187 typedef struct journal_block_tag3_s 188 { 189 __be32 t_blocknr; /* The on-disk block number */ 190 __be32 t_flags; /* See below */ 191 __be32 t_blocknr_high; /* most-significant high 32bits. */ 192 __be32 t_checksum; /* crc32c(uuid+seq+block) */ 193 } journal_block_tag3_t; 194 195 typedef struct journal_block_tag_s 196 { 197 __be32 t_blocknr; /* The on-disk block number */ 198 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ 199 __be16 t_flags; /* See below */ 200 __be32 t_blocknr_high; /* most-significant high 32bits. */ 201 } journal_block_tag_t; 202 203 /* Tail of descriptor or revoke block, for checksumming */ 204 struct jbd2_journal_block_tail { 205 __be32 t_checksum; /* crc32c(uuid+descr_block) */ 206 }; 207 208 /* 209 * The revoke descriptor: used on disk to describe a series of blocks to 210 * be revoked from the log 211 */ 212 typedef struct jbd2_journal_revoke_header_s 213 { 214 journal_header_t r_header; 215 __be32 r_count; /* Count of bytes used in the block */ 216 } jbd2_journal_revoke_header_t; 217 218 /* Definitions for the journal tag flags word: */ 219 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ 220 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ 221 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ 222 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ 223 224 225 /* 226 * The journal superblock. All fields are in big-endian byte order. 227 */ 228 typedef struct journal_superblock_s 229 { 230 /* 0x0000 */ 231 journal_header_t s_header; 232 233 /* 0x000C */ 234 /* Static information describing the journal */ 235 __be32 s_blocksize; /* journal device blocksize */ 236 __be32 s_maxlen; /* total blocks in journal file */ 237 __be32 s_first; /* first block of log information */ 238 239 /* 0x0018 */ 240 /* Dynamic information describing the current state of the log */ 241 __be32 s_sequence; /* first commit ID expected in log */ 242 __be32 s_start; /* blocknr of start of log */ 243 244 /* 0x0020 */ 245 /* Error value, as set by jbd2_journal_abort(). */ 246 __be32 s_errno; 247 248 /* 0x0024 */ 249 /* Remaining fields are only valid in a version-2 superblock */ 250 __be32 s_feature_compat; /* compatible feature set */ 251 __be32 s_feature_incompat; /* incompatible feature set */ 252 __be32 s_feature_ro_compat; /* readonly-compatible feature set */ 253 /* 0x0030 */ 254 __u8 s_uuid[16]; /* 128-bit uuid for journal */ 255 256 /* 0x0040 */ 257 __be32 s_nr_users; /* Nr of filesystems sharing log */ 258 259 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ 260 261 /* 0x0048 */ 262 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ 263 __be32 s_max_trans_data; /* Limit of data blocks per trans. */ 264 265 /* 0x0050 */ 266 __u8 s_checksum_type; /* checksum type */ 267 __u8 s_padding2[3]; 268 __u32 s_padding[42]; 269 __be32 s_checksum; /* crc32c(superblock) */ 270 271 /* 0x0100 */ 272 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ 273 /* 0x0400 */ 274 } journal_superblock_t; 275 276 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */ 277 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ 278 ((j)->j_format_version >= 2 && \ 279 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) 280 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ 281 ((j)->j_format_version >= 2 && \ 282 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) 283 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ 284 ((j)->j_format_version >= 2 && \ 285 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) 286 287 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 288 289 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 290 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 291 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 292 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 293 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 294 295 /* See "journal feature predicate functions" below */ 296 297 /* Features known to this kernel version: */ 298 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM 299 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 300 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ 301 JBD2_FEATURE_INCOMPAT_64BIT | \ 302 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ 303 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ 304 JBD2_FEATURE_INCOMPAT_CSUM_V3) 305 306 #ifdef __KERNEL__ 307 308 #include <linux/fs.h> 309 #include <linux/sched.h> 310 311 enum jbd_state_bits { 312 BH_JBD /* Has an attached ext3 journal_head */ 313 = BH_PrivateStart, 314 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ 315 BH_Freed, /* Has been freed (truncated) */ 316 BH_Revoked, /* Has been revoked from the log */ 317 BH_RevokeValid, /* Revoked flag is valid */ 318 BH_JBDDirty, /* Is dirty but journaled */ 319 BH_State, /* Pins most journal_head state */ 320 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ 321 BH_Shadow, /* IO on shadow buffer is running */ 322 BH_Verified, /* Metadata block has been verified ok */ 323 BH_JBDPrivateStart, /* First bit available for private use by FS */ 324 }; 325 326 BUFFER_FNS(JBD, jbd) 327 BUFFER_FNS(JWrite, jwrite) 328 BUFFER_FNS(JBDDirty, jbddirty) 329 TAS_BUFFER_FNS(JBDDirty, jbddirty) 330 BUFFER_FNS(Revoked, revoked) 331 TAS_BUFFER_FNS(Revoked, revoked) 332 BUFFER_FNS(RevokeValid, revokevalid) 333 TAS_BUFFER_FNS(RevokeValid, revokevalid) 334 BUFFER_FNS(Freed, freed) 335 BUFFER_FNS(Shadow, shadow) 336 BUFFER_FNS(Verified, verified) 337 338 static inline struct buffer_head *jh2bh(struct journal_head *jh) 339 { 340 return jh->b_bh; 341 } 342 343 static inline struct journal_head *bh2jh(struct buffer_head *bh) 344 { 345 return bh->b_private; 346 } 347 348 static inline void jbd_lock_bh_state(struct buffer_head *bh) 349 { 350 bit_spin_lock(BH_State, &bh->b_state); 351 } 352 353 static inline int jbd_trylock_bh_state(struct buffer_head *bh) 354 { 355 return bit_spin_trylock(BH_State, &bh->b_state); 356 } 357 358 static inline int jbd_is_locked_bh_state(struct buffer_head *bh) 359 { 360 return bit_spin_is_locked(BH_State, &bh->b_state); 361 } 362 363 static inline void jbd_unlock_bh_state(struct buffer_head *bh) 364 { 365 bit_spin_unlock(BH_State, &bh->b_state); 366 } 367 368 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) 369 { 370 bit_spin_lock(BH_JournalHead, &bh->b_state); 371 } 372 373 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) 374 { 375 bit_spin_unlock(BH_JournalHead, &bh->b_state); 376 } 377 378 #define J_ASSERT(assert) BUG_ON(!(assert)) 379 380 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) 381 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) 382 383 #if defined(JBD2_PARANOID_IOFAIL) 384 #define J_EXPECT(expr, why...) J_ASSERT(expr) 385 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) 386 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) 387 #else 388 #define __journal_expect(expr, why...) \ 389 ({ \ 390 int val = (expr); \ 391 if (!val) { \ 392 printk(KERN_ERR \ 393 "JBD2 unexpected failure: %s: %s;\n", \ 394 __func__, #expr); \ 395 printk(KERN_ERR why "\n"); \ 396 } \ 397 val; \ 398 }) 399 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) 400 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) 401 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) 402 #endif 403 404 /* Flags in jbd_inode->i_flags */ 405 #define __JI_COMMIT_RUNNING 0 406 /* Commit of the inode data in progress. We use this flag to protect us from 407 * concurrent deletion of inode. We cannot use reference to inode for this 408 * since we cannot afford doing last iput() on behalf of kjournald 409 */ 410 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) 411 412 /** 413 * struct jbd_inode is the structure linking inodes in ordered mode 414 * present in a transaction so that we can sync them during commit. 415 */ 416 struct jbd2_inode { 417 /* Which transaction does this inode belong to? Either the running 418 * transaction or the committing one. [j_list_lock] */ 419 transaction_t *i_transaction; 420 421 /* Pointer to the running transaction modifying inode's data in case 422 * there is already a committing transaction touching it. [j_list_lock] */ 423 transaction_t *i_next_transaction; 424 425 /* List of inodes in the i_transaction [j_list_lock] */ 426 struct list_head i_list; 427 428 /* VFS inode this inode belongs to [constant during the lifetime 429 * of the structure] */ 430 struct inode *i_vfs_inode; 431 432 /* Flags of inode [j_list_lock] */ 433 unsigned long i_flags; 434 }; 435 436 struct jbd2_revoke_table_s; 437 438 /** 439 * struct handle_s - The handle_s type is the concrete type associated with 440 * handle_t. 441 * @h_transaction: Which compound transaction is this update a part of? 442 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. 443 * @h_ref: Reference count on this handle 444 * @h_err: Field for caller's use to track errors through large fs operations 445 * @h_sync: flag for sync-on-close 446 * @h_jdata: flag to force data journaling 447 * @h_aborted: flag indicating fatal error on handle 448 **/ 449 450 /* Docbook can't yet cope with the bit fields, but will leave the documentation 451 * in so it can be fixed later. 452 */ 453 454 struct jbd2_journal_handle 455 { 456 union { 457 /* Which compound transaction is this update a part of? */ 458 transaction_t *h_transaction; 459 /* Which journal handle belongs to - used iff h_reserved set */ 460 journal_t *h_journal; 461 }; 462 463 /* Handle reserved for finishing the logical operation */ 464 handle_t *h_rsv_handle; 465 466 /* Number of remaining buffers we are allowed to dirty: */ 467 int h_buffer_credits; 468 469 /* Reference count on this handle */ 470 int h_ref; 471 472 /* Field for caller's use to track errors through large fs */ 473 /* operations */ 474 int h_err; 475 476 /* Flags [no locking] */ 477 unsigned int h_sync: 1; /* sync-on-close */ 478 unsigned int h_jdata: 1; /* force data journaling */ 479 unsigned int h_reserved: 1; /* handle with reserved credits */ 480 unsigned int h_aborted: 1; /* fatal error on handle */ 481 unsigned int h_type: 8; /* for handle statistics */ 482 unsigned int h_line_no: 16; /* for handle statistics */ 483 484 unsigned long h_start_jiffies; 485 unsigned int h_requested_credits; 486 487 #ifdef CONFIG_DEBUG_LOCK_ALLOC 488 struct lockdep_map h_lockdep_map; 489 #endif 490 }; 491 492 493 /* 494 * Some stats for checkpoint phase 495 */ 496 struct transaction_chp_stats_s { 497 unsigned long cs_chp_time; 498 __u32 cs_forced_to_close; 499 __u32 cs_written; 500 __u32 cs_dropped; 501 }; 502 503 /* The transaction_t type is the guts of the journaling mechanism. It 504 * tracks a compound transaction through its various states: 505 * 506 * RUNNING: accepting new updates 507 * LOCKED: Updates still running but we don't accept new ones 508 * RUNDOWN: Updates are tidying up but have finished requesting 509 * new buffers to modify (state not used for now) 510 * FLUSH: All updates complete, but we are still writing to disk 511 * COMMIT: All data on disk, writing commit record 512 * FINISHED: We still have to keep the transaction for checkpointing. 513 * 514 * The transaction keeps track of all of the buffers modified by a 515 * running transaction, and all of the buffers committed but not yet 516 * flushed to home for finished transactions. 517 */ 518 519 /* 520 * Lock ranking: 521 * 522 * j_list_lock 523 * ->jbd_lock_bh_journal_head() (This is "innermost") 524 * 525 * j_state_lock 526 * ->jbd_lock_bh_state() 527 * 528 * jbd_lock_bh_state() 529 * ->j_list_lock 530 * 531 * j_state_lock 532 * ->t_handle_lock 533 * 534 * j_state_lock 535 * ->j_list_lock (journal_unmap_buffer) 536 * 537 */ 538 539 struct transaction_s 540 { 541 /* Pointer to the journal for this transaction. [no locking] */ 542 journal_t *t_journal; 543 544 /* Sequence number for this transaction [no locking] */ 545 tid_t t_tid; 546 547 /* 548 * Transaction's current state 549 * [no locking - only kjournald2 alters this] 550 * [j_list_lock] guards transition of a transaction into T_FINISHED 551 * state and subsequent call of __jbd2_journal_drop_transaction() 552 * FIXME: needs barriers 553 * KLUDGE: [use j_state_lock] 554 */ 555 enum { 556 T_RUNNING, 557 T_LOCKED, 558 T_FLUSH, 559 T_COMMIT, 560 T_COMMIT_DFLUSH, 561 T_COMMIT_JFLUSH, 562 T_COMMIT_CALLBACK, 563 T_FINISHED 564 } t_state; 565 566 /* 567 * Where in the log does this transaction's commit start? [no locking] 568 */ 569 unsigned long t_log_start; 570 571 /* Number of buffers on the t_buffers list [j_list_lock] */ 572 int t_nr_buffers; 573 574 /* 575 * Doubly-linked circular list of all buffers reserved but not yet 576 * modified by this transaction [j_list_lock] 577 */ 578 struct journal_head *t_reserved_list; 579 580 /* 581 * Doubly-linked circular list of all metadata buffers owned by this 582 * transaction [j_list_lock] 583 */ 584 struct journal_head *t_buffers; 585 586 /* 587 * Doubly-linked circular list of all forget buffers (superseded 588 * buffers which we can un-checkpoint once this transaction commits) 589 * [j_list_lock] 590 */ 591 struct journal_head *t_forget; 592 593 /* 594 * Doubly-linked circular list of all buffers still to be flushed before 595 * this transaction can be checkpointed. [j_list_lock] 596 */ 597 struct journal_head *t_checkpoint_list; 598 599 /* 600 * Doubly-linked circular list of all buffers submitted for IO while 601 * checkpointing. [j_list_lock] 602 */ 603 struct journal_head *t_checkpoint_io_list; 604 605 /* 606 * Doubly-linked circular list of metadata buffers being shadowed by log 607 * IO. The IO buffers on the iobuf list and the shadow buffers on this 608 * list match each other one for one at all times. [j_list_lock] 609 */ 610 struct journal_head *t_shadow_list; 611 612 /* 613 * List of inodes whose data we've modified in data=ordered mode. 614 * [j_list_lock] 615 */ 616 struct list_head t_inode_list; 617 618 /* 619 * Protects info related to handles 620 */ 621 spinlock_t t_handle_lock; 622 623 /* 624 * Longest time some handle had to wait for running transaction 625 */ 626 unsigned long t_max_wait; 627 628 /* 629 * When transaction started 630 */ 631 unsigned long t_start; 632 633 /* 634 * When commit was requested 635 */ 636 unsigned long t_requested; 637 638 /* 639 * Checkpointing stats [j_checkpoint_sem] 640 */ 641 struct transaction_chp_stats_s t_chp_stats; 642 643 /* 644 * Number of outstanding updates running on this transaction 645 * [t_handle_lock] 646 */ 647 atomic_t t_updates; 648 649 /* 650 * Number of buffers reserved for use by all handles in this transaction 651 * handle but not yet modified. [t_handle_lock] 652 */ 653 atomic_t t_outstanding_credits; 654 655 /* 656 * Forward and backward links for the circular list of all transactions 657 * awaiting checkpoint. [j_list_lock] 658 */ 659 transaction_t *t_cpnext, *t_cpprev; 660 661 /* 662 * When will the transaction expire (become due for commit), in jiffies? 663 * [no locking] 664 */ 665 unsigned long t_expires; 666 667 /* 668 * When this transaction started, in nanoseconds [no locking] 669 */ 670 ktime_t t_start_time; 671 672 /* 673 * How many handles used this transaction? [t_handle_lock] 674 */ 675 atomic_t t_handle_count; 676 677 /* 678 * This transaction is being forced and some process is 679 * waiting for it to finish. 680 */ 681 unsigned int t_synchronous_commit:1; 682 683 /* Disk flush needs to be sent to fs partition [no locking] */ 684 int t_need_data_flush; 685 686 /* 687 * For use by the filesystem to store fs-specific data 688 * structures associated with the transaction 689 */ 690 struct list_head t_private_list; 691 }; 692 693 struct transaction_run_stats_s { 694 unsigned long rs_wait; 695 unsigned long rs_request_delay; 696 unsigned long rs_running; 697 unsigned long rs_locked; 698 unsigned long rs_flushing; 699 unsigned long rs_logging; 700 701 __u32 rs_handle_count; 702 __u32 rs_blocks; 703 __u32 rs_blocks_logged; 704 }; 705 706 struct transaction_stats_s { 707 unsigned long ts_tid; 708 unsigned long ts_requested; 709 struct transaction_run_stats_s run; 710 }; 711 712 static inline unsigned long 713 jbd2_time_diff(unsigned long start, unsigned long end) 714 { 715 if (end >= start) 716 return end - start; 717 718 return end + (MAX_JIFFY_OFFSET - start); 719 } 720 721 #define JBD2_NR_BATCH 64 722 723 /** 724 * struct journal_s - The journal_s type is the concrete type associated with 725 * journal_t. 726 * @j_flags: General journaling state flags 727 * @j_errno: Is there an outstanding uncleared error on the journal (from a 728 * prior abort)? 729 * @j_sb_buffer: First part of superblock buffer 730 * @j_superblock: Second part of superblock buffer 731 * @j_format_version: Version of the superblock format 732 * @j_state_lock: Protect the various scalars in the journal 733 * @j_barrier_count: Number of processes waiting to create a barrier lock 734 * @j_barrier: The barrier lock itself 735 * @j_running_transaction: The current running transaction.. 736 * @j_committing_transaction: the transaction we are pushing to disk 737 * @j_checkpoint_transactions: a linked circular list of all transactions 738 * waiting for checkpointing 739 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction 740 * to start committing, or for a barrier lock to be released 741 * @j_wait_done_commit: Wait queue for waiting for commit to complete 742 * @j_wait_commit: Wait queue to trigger commit 743 * @j_wait_updates: Wait queue to wait for updates to complete 744 * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop 745 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints 746 * @j_head: Journal head - identifies the first unused block in the journal 747 * @j_tail: Journal tail - identifies the oldest still-used block in the 748 * journal. 749 * @j_free: Journal free - how many free blocks are there in the journal? 750 * @j_first: The block number of the first usable block 751 * @j_last: The block number one beyond the last usable block 752 * @j_dev: Device where we store the journal 753 * @j_blocksize: blocksize for the location where we store the journal. 754 * @j_blk_offset: starting block offset for into the device where we store the 755 * journal 756 * @j_fs_dev: Device which holds the client fs. For internal journal this will 757 * be equal to j_dev 758 * @j_reserved_credits: Number of buffers reserved from the running transaction 759 * @j_maxlen: Total maximum capacity of the journal region on disk. 760 * @j_list_lock: Protects the buffer lists and internal buffer state. 761 * @j_inode: Optional inode where we store the journal. If present, all journal 762 * block numbers are mapped into this inode via bmap(). 763 * @j_tail_sequence: Sequence number of the oldest transaction in the log 764 * @j_transaction_sequence: Sequence number of the next transaction to grant 765 * @j_commit_sequence: Sequence number of the most recently committed 766 * transaction 767 * @j_commit_request: Sequence number of the most recent transaction wanting 768 * commit 769 * @j_uuid: Uuid of client object. 770 * @j_task: Pointer to the current commit thread for this journal 771 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a 772 * single compound commit transaction 773 * @j_commit_interval: What is the maximum transaction lifetime before we begin 774 * a commit? 775 * @j_commit_timer: The timer used to wakeup the commit thread 776 * @j_revoke_lock: Protect the revoke table 777 * @j_revoke: The revoke table - maintains the list of revoked blocks in the 778 * current transaction. 779 * @j_revoke_table: alternate revoke tables for j_revoke 780 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction 781 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the 782 * number that will fit in j_blocksize 783 * @j_last_sync_writer: most recent pid which did a synchronous write 784 * @j_history: Buffer storing the transactions statistics history 785 * @j_history_max: Maximum number of transactions in the statistics history 786 * @j_history_cur: Current number of transactions in the statistics history 787 * @j_history_lock: Protect the transactions statistics history 788 * @j_proc_entry: procfs entry for the jbd statistics directory 789 * @j_stats: Overall statistics 790 * @j_private: An opaque pointer to fs-private information. 791 */ 792 793 struct journal_s 794 { 795 /* General journaling state flags [j_state_lock] */ 796 unsigned long j_flags; 797 798 /* 799 * Is there an outstanding uncleared error on the journal (from a prior 800 * abort)? [j_state_lock] 801 */ 802 int j_errno; 803 804 /* The superblock buffer */ 805 struct buffer_head *j_sb_buffer; 806 journal_superblock_t *j_superblock; 807 808 /* Version of the superblock format */ 809 int j_format_version; 810 811 /* 812 * Protect the various scalars in the journal 813 */ 814 rwlock_t j_state_lock; 815 816 /* 817 * Number of processes waiting to create a barrier lock [j_state_lock] 818 */ 819 int j_barrier_count; 820 821 /* The barrier lock itself */ 822 struct mutex j_barrier; 823 824 /* 825 * Transactions: The current running transaction... 826 * [j_state_lock] [caller holding open handle] 827 */ 828 transaction_t *j_running_transaction; 829 830 /* 831 * the transaction we are pushing to disk 832 * [j_state_lock] [caller holding open handle] 833 */ 834 transaction_t *j_committing_transaction; 835 836 /* 837 * ... and a linked circular list of all transactions waiting for 838 * checkpointing. [j_list_lock] 839 */ 840 transaction_t *j_checkpoint_transactions; 841 842 /* 843 * Wait queue for waiting for a locked transaction to start committing, 844 * or for a barrier lock to be released 845 */ 846 wait_queue_head_t j_wait_transaction_locked; 847 848 /* Wait queue for waiting for commit to complete */ 849 wait_queue_head_t j_wait_done_commit; 850 851 /* Wait queue to trigger commit */ 852 wait_queue_head_t j_wait_commit; 853 854 /* Wait queue to wait for updates to complete */ 855 wait_queue_head_t j_wait_updates; 856 857 /* Wait queue to wait for reserved buffer credits to drop */ 858 wait_queue_head_t j_wait_reserved; 859 860 /* Semaphore for locking against concurrent checkpoints */ 861 struct mutex j_checkpoint_mutex; 862 863 /* 864 * List of buffer heads used by the checkpoint routine. This 865 * was moved from jbd2_log_do_checkpoint() to reduce stack 866 * usage. Access to this array is controlled by the 867 * j_checkpoint_mutex. [j_checkpoint_mutex] 868 */ 869 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; 870 871 /* 872 * Journal head: identifies the first unused block in the journal. 873 * [j_state_lock] 874 */ 875 unsigned long j_head; 876 877 /* 878 * Journal tail: identifies the oldest still-used block in the journal. 879 * [j_state_lock] 880 */ 881 unsigned long j_tail; 882 883 /* 884 * Journal free: how many free blocks are there in the journal? 885 * [j_state_lock] 886 */ 887 unsigned long j_free; 888 889 /* 890 * Journal start and end: the block numbers of the first usable block 891 * and one beyond the last usable block in the journal. [j_state_lock] 892 */ 893 unsigned long j_first; 894 unsigned long j_last; 895 896 /* 897 * Device, blocksize and starting block offset for the location where we 898 * store the journal. 899 */ 900 struct block_device *j_dev; 901 int j_blocksize; 902 unsigned long long j_blk_offset; 903 char j_devname[BDEVNAME_SIZE+24]; 904 905 /* 906 * Device which holds the client fs. For internal journal this will be 907 * equal to j_dev. 908 */ 909 struct block_device *j_fs_dev; 910 911 /* Total maximum capacity of the journal region on disk. */ 912 unsigned int j_maxlen; 913 914 /* Number of buffers reserved from the running transaction */ 915 atomic_t j_reserved_credits; 916 917 /* 918 * Protects the buffer lists and internal buffer state. 919 */ 920 spinlock_t j_list_lock; 921 922 /* Optional inode where we store the journal. If present, all */ 923 /* journal block numbers are mapped into this inode via */ 924 /* bmap(). */ 925 struct inode *j_inode; 926 927 /* 928 * Sequence number of the oldest transaction in the log [j_state_lock] 929 */ 930 tid_t j_tail_sequence; 931 932 /* 933 * Sequence number of the next transaction to grant [j_state_lock] 934 */ 935 tid_t j_transaction_sequence; 936 937 /* 938 * Sequence number of the most recently committed transaction 939 * [j_state_lock]. 940 */ 941 tid_t j_commit_sequence; 942 943 /* 944 * Sequence number of the most recent transaction wanting commit 945 * [j_state_lock] 946 */ 947 tid_t j_commit_request; 948 949 /* 950 * Journal uuid: identifies the object (filesystem, LVM volume etc) 951 * backed by this journal. This will eventually be replaced by an array 952 * of uuids, allowing us to index multiple devices within a single 953 * journal and to perform atomic updates across them. 954 */ 955 __u8 j_uuid[16]; 956 957 /* Pointer to the current commit thread for this journal */ 958 struct task_struct *j_task; 959 960 /* 961 * Maximum number of metadata buffers to allow in a single compound 962 * commit transaction 963 */ 964 int j_max_transaction_buffers; 965 966 /* 967 * What is the maximum transaction lifetime before we begin a commit? 968 */ 969 unsigned long j_commit_interval; 970 971 /* The timer used to wakeup the commit thread: */ 972 struct timer_list j_commit_timer; 973 974 /* 975 * The revoke table: maintains the list of revoked blocks in the 976 * current transaction. [j_revoke_lock] 977 */ 978 spinlock_t j_revoke_lock; 979 struct jbd2_revoke_table_s *j_revoke; 980 struct jbd2_revoke_table_s *j_revoke_table[2]; 981 982 /* 983 * array of bhs for jbd2_journal_commit_transaction 984 */ 985 struct buffer_head **j_wbuf; 986 int j_wbufsize; 987 988 /* 989 * this is the pid of hte last person to run a synchronous operation 990 * through the journal 991 */ 992 pid_t j_last_sync_writer; 993 994 /* 995 * the average amount of time in nanoseconds it takes to commit a 996 * transaction to disk. [j_state_lock] 997 */ 998 u64 j_average_commit_time; 999 1000 /* 1001 * minimum and maximum times that we should wait for 1002 * additional filesystem operations to get batched into a 1003 * synchronous handle in microseconds 1004 */ 1005 u32 j_min_batch_time; 1006 u32 j_max_batch_time; 1007 1008 /* This function is called when a transaction is closed */ 1009 void (*j_commit_callback)(journal_t *, 1010 transaction_t *); 1011 1012 /* 1013 * Journal statistics 1014 */ 1015 spinlock_t j_history_lock; 1016 struct proc_dir_entry *j_proc_entry; 1017 struct transaction_stats_s j_stats; 1018 1019 /* Failed journal commit ID */ 1020 unsigned int j_failed_commit; 1021 1022 /* 1023 * An opaque pointer to fs-private information. ext3 puts its 1024 * superblock pointer here 1025 */ 1026 void *j_private; 1027 1028 /* Reference to checksum algorithm driver via cryptoapi */ 1029 struct crypto_shash *j_chksum_driver; 1030 1031 /* Precomputed journal UUID checksum for seeding other checksums */ 1032 __u32 j_csum_seed; 1033 }; 1034 1035 /* journal feature predicate functions */ 1036 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \ 1037 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1038 { \ 1039 return ((j)->j_format_version >= 2 && \ 1040 ((j)->j_superblock->s_feature_compat & \ 1041 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \ 1042 } \ 1043 static inline void jbd2_set_feature_##name(journal_t *j) \ 1044 { \ 1045 (j)->j_superblock->s_feature_compat |= \ 1046 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1047 } \ 1048 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1049 { \ 1050 (j)->j_superblock->s_feature_compat &= \ 1051 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \ 1052 } 1053 1054 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \ 1055 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1056 { \ 1057 return ((j)->j_format_version >= 2 && \ 1058 ((j)->j_superblock->s_feature_ro_compat & \ 1059 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \ 1060 } \ 1061 static inline void jbd2_set_feature_##name(journal_t *j) \ 1062 { \ 1063 (j)->j_superblock->s_feature_ro_compat |= \ 1064 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1065 } \ 1066 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1067 { \ 1068 (j)->j_superblock->s_feature_ro_compat &= \ 1069 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \ 1070 } 1071 1072 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \ 1073 static inline bool jbd2_has_feature_##name(journal_t *j) \ 1074 { \ 1075 return ((j)->j_format_version >= 2 && \ 1076 ((j)->j_superblock->s_feature_incompat & \ 1077 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \ 1078 } \ 1079 static inline void jbd2_set_feature_##name(journal_t *j) \ 1080 { \ 1081 (j)->j_superblock->s_feature_incompat |= \ 1082 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1083 } \ 1084 static inline void jbd2_clear_feature_##name(journal_t *j) \ 1085 { \ 1086 (j)->j_superblock->s_feature_incompat &= \ 1087 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \ 1088 } 1089 1090 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM) 1091 1092 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE) 1093 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT) 1094 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT) 1095 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2) 1096 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3) 1097 1098 /* 1099 * Journal flag definitions 1100 */ 1101 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ 1102 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ 1103 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ 1104 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ 1105 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ 1106 #define JBD2_BARRIER 0x020 /* Use IDE barriers */ 1107 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file 1108 * data write error in ordered 1109 * mode */ 1110 #define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */ 1111 1112 /* 1113 * Function declarations for the journaling transaction and buffer 1114 * management 1115 */ 1116 1117 /* Filing buffers */ 1118 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); 1119 extern void __jbd2_journal_refile_buffer(struct journal_head *); 1120 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); 1121 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1122 extern void __journal_free_buffer(struct journal_head *bh); 1123 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); 1124 extern void __journal_clean_data_list(transaction_t *transaction); 1125 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) 1126 { 1127 list_add_tail(&bh->b_assoc_buffers, head); 1128 } 1129 static inline void jbd2_unfile_log_bh(struct buffer_head *bh) 1130 { 1131 list_del_init(&bh->b_assoc_buffers); 1132 } 1133 1134 /* Log buffer allocation */ 1135 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int); 1136 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *); 1137 int jbd2_journal_next_log_block(journal_t *, unsigned long long *); 1138 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, 1139 unsigned long *block); 1140 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1141 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); 1142 1143 /* Commit management */ 1144 extern void jbd2_journal_commit_transaction(journal_t *); 1145 1146 /* Checkpoint list management */ 1147 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy); 1148 int __jbd2_journal_remove_checkpoint(struct journal_head *); 1149 void jbd2_journal_destroy_checkpoint(journal_t *journal); 1150 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); 1151 1152 1153 /* 1154 * Triggers 1155 */ 1156 1157 struct jbd2_buffer_trigger_type { 1158 /* 1159 * Fired a the moment data to write to the journal are known to be 1160 * stable - so either at the moment b_frozen_data is created or just 1161 * before a buffer is written to the journal. mapped_data is a mapped 1162 * buffer that is the frozen data for commit. 1163 */ 1164 void (*t_frozen)(struct jbd2_buffer_trigger_type *type, 1165 struct buffer_head *bh, void *mapped_data, 1166 size_t size); 1167 1168 /* 1169 * Fired during journal abort for dirty buffers that will not be 1170 * committed. 1171 */ 1172 void (*t_abort)(struct jbd2_buffer_trigger_type *type, 1173 struct buffer_head *bh); 1174 }; 1175 1176 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, 1177 void *mapped_data, 1178 struct jbd2_buffer_trigger_type *triggers); 1179 extern void jbd2_buffer_abort_trigger(struct journal_head *jh, 1180 struct jbd2_buffer_trigger_type *triggers); 1181 1182 /* Buffer IO */ 1183 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, 1184 struct journal_head *jh_in, 1185 struct buffer_head **bh_out, 1186 sector_t blocknr); 1187 1188 /* Transaction locking */ 1189 extern void __wait_on_journal (journal_t *); 1190 1191 /* Transaction cache support */ 1192 extern void jbd2_journal_destroy_transaction_cache(void); 1193 extern int jbd2_journal_init_transaction_cache(void); 1194 extern void jbd2_journal_free_transaction(transaction_t *); 1195 1196 /* 1197 * Journal locking. 1198 * 1199 * We need to lock the journal during transaction state changes so that nobody 1200 * ever tries to take a handle on the running transaction while we are in the 1201 * middle of moving it to the commit phase. j_state_lock does this. 1202 * 1203 * Note that the locking is completely interrupt unsafe. We never touch 1204 * journal structures from interrupts. 1205 */ 1206 1207 static inline handle_t *journal_current_handle(void) 1208 { 1209 return current->journal_info; 1210 } 1211 1212 /* The journaling code user interface: 1213 * 1214 * Create and destroy handles 1215 * Register buffer modifications against the current transaction. 1216 */ 1217 1218 extern handle_t *jbd2_journal_start(journal_t *, int nblocks); 1219 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, 1220 gfp_t gfp_mask, unsigned int type, 1221 unsigned int line_no); 1222 extern int jbd2_journal_restart(handle_t *, int nblocks); 1223 extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask); 1224 extern int jbd2_journal_start_reserved(handle_t *handle, 1225 unsigned int type, unsigned int line_no); 1226 extern void jbd2_journal_free_reserved(handle_t *handle); 1227 extern int jbd2_journal_extend (handle_t *, int nblocks); 1228 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); 1229 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); 1230 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); 1231 void jbd2_journal_set_triggers(struct buffer_head *, 1232 struct jbd2_buffer_trigger_type *type); 1233 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); 1234 extern int jbd2_journal_forget (handle_t *, struct buffer_head *); 1235 extern void journal_sync_buffer (struct buffer_head *); 1236 extern int jbd2_journal_invalidatepage(journal_t *, 1237 struct page *, unsigned int, unsigned int); 1238 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); 1239 extern int jbd2_journal_stop(handle_t *); 1240 extern int jbd2_journal_flush (journal_t *); 1241 extern void jbd2_journal_lock_updates (journal_t *); 1242 extern void jbd2_journal_unlock_updates (journal_t *); 1243 1244 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, 1245 struct block_device *fs_dev, 1246 unsigned long long start, int len, int bsize); 1247 extern journal_t * jbd2_journal_init_inode (struct inode *); 1248 extern int jbd2_journal_update_format (journal_t *); 1249 extern int jbd2_journal_check_used_features 1250 (journal_t *, unsigned long, unsigned long, unsigned long); 1251 extern int jbd2_journal_check_available_features 1252 (journal_t *, unsigned long, unsigned long, unsigned long); 1253 extern int jbd2_journal_set_features 1254 (journal_t *, unsigned long, unsigned long, unsigned long); 1255 extern void jbd2_journal_clear_features 1256 (journal_t *, unsigned long, unsigned long, unsigned long); 1257 extern int jbd2_journal_load (journal_t *journal); 1258 extern int jbd2_journal_destroy (journal_t *); 1259 extern int jbd2_journal_recover (journal_t *journal); 1260 extern int jbd2_journal_wipe (journal_t *, int); 1261 extern int jbd2_journal_skip_recovery (journal_t *); 1262 extern void jbd2_journal_update_sb_errno(journal_t *); 1263 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t, 1264 unsigned long, int); 1265 extern void __jbd2_journal_abort_hard (journal_t *); 1266 extern void jbd2_journal_abort (journal_t *, int); 1267 extern int jbd2_journal_errno (journal_t *); 1268 extern void jbd2_journal_ack_err (journal_t *); 1269 extern int jbd2_journal_clear_err (journal_t *); 1270 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); 1271 extern int jbd2_journal_force_commit(journal_t *); 1272 extern int jbd2_journal_force_commit_nested(journal_t *); 1273 extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode); 1274 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, 1275 struct jbd2_inode *inode, loff_t new_size); 1276 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); 1277 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); 1278 1279 /* 1280 * journal_head management 1281 */ 1282 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); 1283 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); 1284 void jbd2_journal_put_journal_head(struct journal_head *jh); 1285 1286 /* 1287 * handle management 1288 */ 1289 extern struct kmem_cache *jbd2_handle_cache; 1290 1291 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) 1292 { 1293 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); 1294 } 1295 1296 static inline void jbd2_free_handle(handle_t *handle) 1297 { 1298 kmem_cache_free(jbd2_handle_cache, handle); 1299 } 1300 1301 /* 1302 * jbd2_inode management (optional, for those file systems that want to use 1303 * dynamically allocated jbd2_inode structures) 1304 */ 1305 extern struct kmem_cache *jbd2_inode_cache; 1306 1307 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) 1308 { 1309 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); 1310 } 1311 1312 static inline void jbd2_free_inode(struct jbd2_inode *jinode) 1313 { 1314 kmem_cache_free(jbd2_inode_cache, jinode); 1315 } 1316 1317 /* Primary revoke support */ 1318 #define JOURNAL_REVOKE_DEFAULT_HASH 256 1319 extern int jbd2_journal_init_revoke(journal_t *, int); 1320 extern void jbd2_journal_destroy_revoke_caches(void); 1321 extern int jbd2_journal_init_revoke_caches(void); 1322 1323 extern void jbd2_journal_destroy_revoke(journal_t *); 1324 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); 1325 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); 1326 extern void jbd2_journal_write_revoke_records(transaction_t *transaction, 1327 struct list_head *log_bufs); 1328 1329 /* Recovery revoke support */ 1330 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); 1331 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); 1332 extern void jbd2_journal_clear_revoke(journal_t *); 1333 extern void jbd2_journal_switch_revoke_table(journal_t *journal); 1334 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); 1335 1336 /* 1337 * The log thread user interface: 1338 * 1339 * Request space in the current transaction, and force transaction commit 1340 * transitions on demand. 1341 */ 1342 1343 int jbd2_log_start_commit(journal_t *journal, tid_t tid); 1344 int __jbd2_log_start_commit(journal_t *journal, tid_t tid); 1345 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); 1346 int jbd2_log_wait_commit(journal_t *journal, tid_t tid); 1347 int jbd2_complete_transaction(journal_t *journal, tid_t tid); 1348 int jbd2_log_do_checkpoint(journal_t *journal); 1349 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); 1350 1351 void __jbd2_log_wait_for_space(journal_t *journal); 1352 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); 1353 extern int jbd2_cleanup_journal_tail(journal_t *); 1354 1355 /* 1356 * is_journal_abort 1357 * 1358 * Simple test wrapper function to test the JBD2_ABORT state flag. This 1359 * bit, when set, indicates that we have had a fatal error somewhere, 1360 * either inside the journaling layer or indicated to us by the client 1361 * (eg. ext3), and that we and should not commit any further 1362 * transactions. 1363 */ 1364 1365 static inline int is_journal_aborted(journal_t *journal) 1366 { 1367 return journal->j_flags & JBD2_ABORT; 1368 } 1369 1370 static inline int is_handle_aborted(handle_t *handle) 1371 { 1372 if (handle->h_aborted || !handle->h_transaction) 1373 return 1; 1374 return is_journal_aborted(handle->h_transaction->t_journal); 1375 } 1376 1377 static inline void jbd2_journal_abort_handle(handle_t *handle) 1378 { 1379 handle->h_aborted = 1; 1380 } 1381 1382 #endif /* __KERNEL__ */ 1383 1384 /* Comparison functions for transaction IDs: perform comparisons using 1385 * modulo arithmetic so that they work over sequence number wraps. */ 1386 1387 static inline int tid_gt(tid_t x, tid_t y) 1388 { 1389 int difference = (x - y); 1390 return (difference > 0); 1391 } 1392 1393 static inline int tid_geq(tid_t x, tid_t y) 1394 { 1395 int difference = (x - y); 1396 return (difference >= 0); 1397 } 1398 1399 extern int jbd2_journal_blocks_per_page(struct inode *inode); 1400 extern size_t journal_tag_bytes(journal_t *journal); 1401 1402 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j) 1403 { 1404 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j); 1405 } 1406 1407 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) 1408 { 1409 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) && 1410 journal->j_chksum_driver == NULL); 1411 1412 return journal->j_chksum_driver != NULL; 1413 } 1414 1415 /* 1416 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for 1417 * transaction control blocks. 1418 */ 1419 #define JBD2_CONTROL_BLOCKS_SHIFT 5 1420 1421 /* 1422 * Return the minimum number of blocks which must be free in the journal 1423 * before a new transaction may be started. Must be called under j_state_lock. 1424 */ 1425 static inline int jbd2_space_needed(journal_t *journal) 1426 { 1427 int nblocks = journal->j_max_transaction_buffers; 1428 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT); 1429 } 1430 1431 /* 1432 * Return number of free blocks in the log. Must be called under j_state_lock. 1433 */ 1434 static inline unsigned long jbd2_log_space_left(journal_t *journal) 1435 { 1436 /* Allow for rounding errors */ 1437 unsigned long free = journal->j_free - 32; 1438 1439 if (journal->j_committing_transaction) { 1440 unsigned long committing = atomic_read(&journal-> 1441 j_committing_transaction->t_outstanding_credits); 1442 1443 /* Transaction + control blocks */ 1444 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT); 1445 } 1446 return free; 1447 } 1448 1449 /* 1450 * Definitions which augment the buffer_head layer 1451 */ 1452 1453 /* journaling buffer types */ 1454 #define BJ_None 0 /* Not journaled */ 1455 #define BJ_Metadata 1 /* Normal journaled metadata */ 1456 #define BJ_Forget 2 /* Buffer superseded by this transaction */ 1457 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ 1458 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ 1459 #define BJ_Types 5 1460 1461 extern int jbd_blocks_per_page(struct inode *inode); 1462 1463 /* JBD uses a CRC32 checksum */ 1464 #define JBD_MAX_CHECKSUM_SIZE 4 1465 1466 static inline u32 jbd2_chksum(journal_t *journal, u32 crc, 1467 const void *address, unsigned int length) 1468 { 1469 struct { 1470 struct shash_desc shash; 1471 char ctx[JBD_MAX_CHECKSUM_SIZE]; 1472 } desc; 1473 int err; 1474 1475 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > 1476 JBD_MAX_CHECKSUM_SIZE); 1477 1478 desc.shash.tfm = journal->j_chksum_driver; 1479 desc.shash.flags = 0; 1480 *(u32 *)desc.ctx = crc; 1481 1482 err = crypto_shash_update(&desc.shash, address, length); 1483 BUG_ON(err); 1484 1485 return *(u32 *)desc.ctx; 1486 } 1487 1488 /* Return most recent uncommitted transaction */ 1489 static inline tid_t jbd2_get_latest_transaction(journal_t *journal) 1490 { 1491 tid_t tid; 1492 1493 read_lock(&journal->j_state_lock); 1494 tid = journal->j_commit_request; 1495 if (journal->j_running_transaction) 1496 tid = journal->j_running_transaction->t_tid; 1497 read_unlock(&journal->j_state_lock); 1498 return tid; 1499 } 1500 1501 #ifdef __KERNEL__ 1502 1503 #define buffer_trace_init(bh) do {} while (0) 1504 #define print_buffer_fields(bh) do {} while (0) 1505 #define print_buffer_trace(bh) do {} while (0) 1506 #define BUFFER_TRACE(bh, info) do {} while (0) 1507 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) 1508 #define JBUFFER_TRACE(jh, info) do {} while (0) 1509 1510 #endif /* __KERNEL__ */ 1511 1512 #define EFSBADCRC EBADMSG /* Bad CRC detected */ 1513 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */ 1514 1515 #endif /* _LINUX_JBD2_H */ 1516